Many communication systems modulate electromagnetic signals from baseband to higher frequencies for transmission, and subsequently demodulate those high frequencies back to their original frequency band when they reach the receiver. The original (or baseband) signal, may be, for example: data, voice or video. These baseband signals may be produced by transducers such as microphones or video cameras, be computer generated, or transferred from an electronic storage device. In general, the high transmission frequencies provide longer range and higher capacity channels than baseband signals, and because high frequency RF signals can propagate through the air, they can be used for wireless channels as well as hard wired or fibre channels.
All of these signals are generally referred to as radio frequency (RF) signals, which are electromagnetic signals, that is, waveforms with electrical and magnetic properties within the electromagnetic spectrum normally associated with radio wave propagation. The electromagnetic spectrum was traditionally divided into 26 alphabetically designated bands, however, the International Telecommunication Union (ITU) formally recognizes 12 bands, from 30 Hz to 3000 GHz. New bands, from 3 THz to 3000 THz, are under active consideration for recognition.
Wired communication systems which employ such modulation and demodulation techniques include computer communication systems such as local area networks (LANs), point to point signalling, and wide area networks (WANs) such as the Internet. These networks generally communication data signals over electrical or optical fibre channels. Wireless communication systems which may employ modulation and demodulation include those for public broadcasting such as AM and FM radio, and UHF and VHF television. Private communication systems may include cellular telephone networks, personal paging devices, HF radio systems used by taxi services, microwave backbone networks, interconnected appliances under the Bluetooth standard, and satellite communications. Other wired and wireless systems which use RF modulation and demodulation would be known to those skilled in the art.
One of the current problems in the art, is to develop physically small and inexpensive modulation and demodulation techniques and devices that have good performance characteristics. For cellular telephones, for example, it is desirable to have transmitters and receivers (which may be referred to in combination as a transceiver) which can be fully integrated onto integrated circuits (ICs).
Several attempts at completely integrated transceiver designs have met with limited success. For example, most RF topology typically requires at least two high quality filters that cannot be economically integrated within any modern IC technology. Other RF receiver topologies exist, such as image rejection architectures, which can be completely integrated on a chip, but lack in overall performance. Most receivers use the “super-heterodyne” topology, which provides excellent performance, but does not meet the desired level of integration for modem wireless systems.
Existing transceiver solutions and their associated problems and limitations are summarized below.
1. Super-Heterodyne
The super-heterodyne receiver uses a two-step frequency translation method to convert an RF signal to a baseband signal. FIG. 1 presents a block diagram of a typical super-heterodyne receiver 10. Generally, the mixers labelled M1 12, MI 14, and MQ 16 are used to translate an incoming RF signal to baseband or to some intermediate frequency (IF). The balance of the components amplify the signal being processed and filter noise from it.
More specifically, the RF band pass filter (BPF1) 18 first filters the incoming signal and corruptive noise coming from the antenna 20, attenuating out of band signals and passing the desired signal (note that this band pass filter 18 may also be a duplexer). A low noise amplifier 22 then amplifies the filtered antenna signal, increasing the strength of the RF signal and reducing the noise figure of the receiver 10. The signal is next filtered by another band pass filter (BPF2) 24 usually identified as an image rejection filter. The desired signal, plus residual unwanted signals, then enter mixer M1 12 which multiplies this signal with a periodic sinusoidal signal generated by the local oscillator (LO1) 26. The mixer M1 12 receives the signal from the image rejection filter 24 and causes both a down-conversion and an up-conversion in the frequency domain. Usually, the down-converted portion is retained at the so-called “Intermediate Frequency” (IF).
Generally, a mixer is a circuit or device that accepts as its input two different frequencies and presents at its output:    (a) a signal equal in frequency to the sum of the frequencies of the input signals;    (b) a signal equal in frequency to the difference between the frequencies of the input signals; and    (c) the original input frequencies.
Note that the frequency conversion process causes a second band of frequencies to be superimposed upon the desired signal at the IF frequency. These “image frequencies” are also passed by the band pass filter 24 and corrupt the desired signal. Note also that the typical embodiment of a mixer is a digital switch, which may generate significantly more tones than those described in (a) through (c).
The IF signal is next filtered by a band pass filter (BPF3) 28 typically called the channel filter, which is centered around the IF frequency, thus filtering out mixer signals (a) and (c) above.
The signal is then amplified by an amplifier (IFA) 30, and is split into its in-phase (I) and quadrature (O) components, using mixers MI 14 and MQ 16, and orthogonal mixing signals generated by local oscillator (LO2) 32 and 90 degree phase shifter 34. LO2 32 generates a regular, periodic signal which is typically tuned to the IF frequency, so that the signals coming from the outputs of MI 14 and MQ 16 are now at baseband, that is, the frequency at which they were originally generated. The two signals are next filtered using low pass filters LPFI 36 and LPFQ 38 to remove the unwanted products of the mixing process, producing baseband I and Q signals. The signals may then be amplified by gain-controlled amplifiers AGCI 40 and AGCQ 42, and digitized via analog to digital converters ADI 44 and ADO 46 if required by the receiver.
The main problems with the super-heterodyne design are:                it requires expensive off-chip components, particularly band pass filters 18, 24, 28, and low pass filters 36, 38 to remove unwanted signal components;        the off-chip components require design trade-offs that increase power consumption and reduce system gain;        image rejection is limited by the off-chip components, not by the target integration technology;        isolation from digital noise can be a problem; and        it is not fully integratable.2. Image Rejection Architectures        
Several image rejection architectures exist, but are not widely used. The two most well known being the Hartley Image Rejection Architecture and the Weaver Image Rejection Architecture. There are other designs, which are generally based on these two architectures, and other methods which employ poly-phase filters to cancel image components. Generally, either accurate signal phase shifts or accurate generation of quadrature local oscillators are employed in these architectures to cancel the image frequencies. The amount of image cancellation is directly dependent upon the degree of accuracy in producing the phase shift or in producing the quadrature local oscillator signals.
Although the integratability of these architectures is high, their performance is relatively poor due to the required accuracy of the phase shifts and quadrature oscillators. This architecture has been used for dual mode receivers on a single chip.
3. Direct Conversion
Direct conversion architectures demodulate RF signals to baseband in a single step, by mixing the RF signal with a local oscillator signal at the carrier frequency of the RF signal. There is therefore no image frequency, and no image components to corrupt the signal. Direct conversion receivers offer a high level of integratability, but also have several important problems. Hence, direct conversion receivers have thus far proved useful only for signalling formats that do not place appreciable signal energy near DC after conversion to baseband.
A typical direct conversion receiver is shown in FIG. 2. The RF band pass filter (BPF1) 18 first filters the signal coming from the antenna 20 (this band pass filter 18 may also be a duplexer). A low noise amplifier 22 is then used to amplify the filtered antenna signal, increasing the strength of the RF signal and reducing the noise figure of the receiver 10.
The signal is then split into its quadrature components and demodulated in a single stage using mixers MI 14 and MQ 16, and orthogonal signals generated by local oscillator (LO2) 32 and 90 degree phase shifter 34. L02 32 generates a regular, periodic signal which is tuned to the incoming wanted frequency rather than an IF frequency as in the case of the super-heterodyne receiver. The signals coming from the outputs of MI 14 and MQ 16 are now at baseband, that is, the frequency at which they were originally generated. The two signals are next filtered using low pass filters LPFI 36 and LPFQ 38, are amplified by gain-controlled amplifiers AGCI 40 and AGCQ 42, and are digitized via analog to digital converters ADI 44 and ADQ 46.
Direct conversion RF receivers have several advantages over super-heterodyne systems in term of cost, power, and level of integration, however, there are also several serious problems with direct conversion. These problems include:                noise near baseband (that is, 1/f noise) which corrupts the desired signal;        local oscillator (LO) leakage in the RF path that creates DC offsets. As the LO frequency is the same as the incoming signal being demodulated, any leakage of the LO signal onto the antenna side of the mixer will pass through to the output side as well;        local oscillator leakage into the RF path that causes desensitization. Desensitation is the reduction of desired signal gain as a result of receiver reaction to an undesired signal. The gain reduction is generally due to overload of some portion of the receiver, such as the AGC circuitry, resulting in suppression of the desired signal because the receiver will no longer respond linearly to incremental changes in input voltage.        noise inherent to mixed-signal integrated circuits corrupts the desired signal;        large on-chip capacitors are required to remove unwanted noise and signal energy near DC, which makes integrability expensive. These capacitors are typically placed between the mixers and the low pass filters; and        errors are generated in the quadrature signals due to inaccuracies in the 90 degree phase shifter.4. Near Zero-IF Conversion        
This receiver architecture is similar to the direct conversion architecture, in that the RF input signal band is translated brought close to baseband in a single step using a regular, periodic oscillator signal. However, the desired signal is not brought exactly to baseband and therefore DC offsets and 1/f noise do not contaminate the output signal. Image frequencies are again a problem though, as in the case of the super-heterodyne structure.
Additional problems encountered with near zero-IF architectures include:                the need for very accurate quadrature local oscillators;        the need for several balanced signal paths for purposes of image cancellation;        noise inherent to mixed-signal integrated circuits which corrupts the desired output signal; and        isolation from digital noise can be a problem.5. Sub-Sampling Down-Conversion        
This method of signal down-conversion utilizes subsampling of the input signal to effect the frequency translation, that is, the input signal is sampled at a lower rate than the signal frequency. This may be done, for example, by use of a sample and hold circuit.
Although the level of integration possible with this technique is the highest among those discussed thus far, the subsampling down-conversion method suffers from two major drawbacks:                subsampling of the RF signal causes aliasing of unwanted noise power to DC. Sampling by a factor of m increases the down-converted noise power of the sampling circuit by a factor of 2m; and        subsampling also increases the effect of noise in the sampling clock. In fact, the clock phase noise power is increased by m2 for sampling by a factor of m.        
There is therefore a need for a method and apparatus of modulating and demodulating RF signals which allows the desired integrability along with good performance.